JLE

Médecine de la Reproduction

MENU

Bayesian principles and concepts and applications in medicine Volume 26, issue 1, janvier-février-mars 2024

Figures


  • Figure 1.

  • Figure 2.

  • Figure 3.

  • Figure 4.
Author
GMRC, Pôle de santé publique, santé au travail, hygiène hospitalière, CIC, hôpitaux universitaires de Strasbourg, Strasbourg, France iCUBE, IMR 7357
* Tirés à part : N. Meyer

The medical literature extensively uses statistical methods, particularly p-values obtained after null hypothesis testing. These tools determine whether two proportions differ by dichotomising the p-value on either side of an alpha threshold, typically set at 0.05, based on data collected from a sample of subjects. In addition to these classical methods, Bayesian statistical analyses have emerged in the literature over the past thirty years. These methods differ from traditional ones in that they do not rely on p-values and instead use a subjectivist definition of probability. This results in the expression of an a priori distribution for the parameter to be estimated, which specifies the available knowledge about the parameter before the experiment is carried out. This a priori distribution is combined with the probability of the data observed in the sample to update our knowledge of the parameter of interest (difference in proportion or odds ratio). The principles and concepts underlying Bayesian methods are presented here, along with a few examples of their application, particularly in reproductive medicine, to show the advantages of these methods over conventional methods.